US20110187746A1 - Image Processing Device, Image Processing Method, and Program - Google Patents

Abstract

There is provided an image processing device including: a data storage unit storing feature data indicating a feature of appearance of one or more physical objects; an environment map building unit for building an environment map based on an input image obtained by imaging a real space and the feature data, the environment map representing a position of a physical object present in the real space; a control unit for acquiring procedure data for a set of procedures of operation to be performed in the real space, the procedure data defining a correspondence between a direction for each procedure and position information designating a position at which the direction is to be displayed; and a superimposing unit for generating an output image by superimposing the direction for each procedure at a position in the input image determined based on the environment map and the position information, using the procedure data.

Description

BACKGROUND OF THE INVENTION
• 1. Field of the Invention
• The present invention relates to an image processing device, an image processing method, and a program.
• 1. Description of the Related Art
• Recently, technology called augmented reality (AR) in which an image obtained by imaging a real space is processed and then presented to a user has been receiving attention. In the AR technology, useful information related to a physical object in a real space present in an input image may be inserted into the image and output as an output image, for example. That is, in the AR technology, typically, a large part of the image presented to the user shows the real space, and a part of the image may be processed in accordance with a purpose. Such a characteristic contrasts it with virtual reality in which an entire (or a large part) of the output image is composed using computer graphics (CG). By using the AR technology, for example, advantages such as easy recognition of a situation of the real space by a user or operation support based on the output image may be provided.
• In the AR technology, in order to present actually useful information to the user, it is important that a computer accurately recognize a situation of the real space. Therefore, technology aimed at recognizing the situation of the real space, which serves as a basis of the AR technology, has been developed. For example, Japanese Patent Application Laid-Open Publication No. 2008-304268 discloses a method of dynamically generating an environment map representing a three-dimensional position of physical objects existing in a real space by applying technology called simultaneous localization and mapping (SLAM) capable of simultaneously estimating a position and a posture of a camera and a position of a feature point shown in an image of the camera. Further, a basic principle of the SLAM technology using a monocular camera is disclosed in Andrew J. Davison's, “Real-Time Simultaneous Localization and Mapping with a Single Camera,” Proceedings of the 9th IEEE International Conference on Computer VisionVolume 2, 2003, pp. 1403-1410.
• Now, information terminals capable of performing enhanced information processing are widely used and users use the information terminals to view various information. An example of the information includes directions. For example, various information, such as table manners, a cooking procedure, a method of manipulating or repairing electrical appliances, or the like, as well as directions of electronic devices, describing a procedure of an operation to be performed in a real space are made in an electronic form and provided to the users via web pages or other applications.
SUMMARY OF THE INVENTION
• However, existing electronic directions are generally described based on the premise that the electronic directions are displayed on a two-dimensional screen, like paper directions. Accordingly, some information about positions of operation in a three-dimensional real space is missed, often making it difficult for a user to intuitively understand a procedure of the operation. In particular, for an operation in a real space in which an environment surrounding a user may be dynamically changed, such environment is not described in the directions, making it more difficult to understand the directions.
• Accordingly, if an environment map three-dimensionally representing an environment surrounding a user can be dynamically built and a direction for each procedure of the operation can be displayed in a position associated with the environment map, the operation procedure is expected to be intuitively and easily understood by the user.
• In light of the foregoing, it is desirable to provide an image processing device, an image processing method, and a program which allow an operation procedure performed in a real space to be intuitively and easily understood by applying an environment map.
• According to an embodiment of the present invention, there is provided an image processing device including: a data storage unit having feature data stored therein, the feature data indicating a feature of appearance of one or more physical objects; an environment map building unit for building an environment map based on an input image obtained by imaging a real space using an imaging device and the feature data stored in the data storage unit, the environment map representing a position of a physical object present in the real space; a control unit for acquiring procedure data for a set of procedures of operation to be performed in the real space, the procedure data defining a correspondence between a direction for each procedure and position information designating a position at which the direction is to be displayed; and a superimposing unit for generating an output image by superimposing the direction for each procedure included in the set of procedures at a position in the input image determined based on the environment map and the position information, using the procedure data acquired by the control unit.
• According to such a configuration, the direction for each procedure included in a set of procedures of operation to be performed in the real space is superimposed at a position in the input image determined based on the environment map three-dimensionally representing positions of physical objects present in the real space, according to definition by procedure data.
• The position information may designate a position in the environment map at which the direction is to be displayed, by specifying a physical object related to the direction for each procedure.
• The procedure data may define a further correspondence between the direction for each procedure and a condition for progressing displaying of each direction.
• The condition for progressing displaying of each direction may include a condition according to a position or a posture of a physical object represented by the environment map.
• The control unit may control displaying of the direction for each procedure included in the set of procedures according to the procedure data.
• The image processing device may further include a detection unit for dynamically detecting a position in the real space of the imaging device based on the input image and the feature data, and the position in the input image at which the direction for each procedure is to be superimposed may be determined according to the position in the real space of the imaging device detected by the detection unit.
• According to another embodiment of the present invention, there is provided an image processing method in an image processing device including a storage medium having feature data stored therein, the feature data indicating a feature of appearance of one or more physical objects, the method including the steps of: acquiring an input image obtained by imaging a real space using an imaging device; building an environment map based on the input image and the feature data, the environment map representing a position of a physical object present in the real space; acquiring procedure data for a set of procedures of operation to be performed in the real space, the procedure data defining a correspondence between a direction for each procedure and position information designating a position at which the direction is to be displayed; and generating an output image by superimposing the direction for each procedure included in the set of procedures at a position in the input image determined based on the environment map and the position information, using the acquired procedure data.
• According to another embodiment of the present invention, there is provided a program for causing a computer, which controls an image processing device including a storage medium having feature data stored therein, the feature data indicating a feature of appearance of one or more physical objects, to function as: an environment map building unit for building an environment map based on an input image obtained by imaging a real space using an imaging device and the feature data, the environment map representing a position of a physical object present in the real space; a control unit for acquiring procedure data for a set of procedures of operation to be performed in the real space, the procedure data defining a correspondence between a direction for each procedure and position information designating a position at which the direction is to be displayed; and a superimposing unit for generating an output image by superimposing the direction for each procedure included in the set of procedures at a position in the input image determined based on the environment map and the position information, using the procedure data acquired by the control unit.
• As described above, according to an image processing device, an image processing method, and a program in an embodiment of the present invention, it is possible to enable an operation procedure performed in the real space to be intuitively and easily understood by applying the environment map.
BRIEF DESCRIPTION OF THE DRAWINGS
• FIG. 1A is a schematic diagram showing a first example of an environment in which an image processing device can be used according to an embodiment;
• FIG. 1B is a schematic diagram showing a second example of the environment in which the image processing device can be used according to the embodiment;
• FIG. 2A is a first illustrative diagram for illustrating an image processing device according to an embodiment;
• FIG. 2B is a second illustrative diagram for illustrating an image processing device according to an embodiment;
• FIG. 3A is an illustrative diagram showing an example of an input image that can be acquired in a real environment shown inFIG. 1A;
• FIG. 3B is an illustrative diagram showing an example of an input image that can be acquired in a real environment shown inFIG. 1B;
• FIG. 4 is a block diagram showing an example of a configuration of an image processing device according to an embodiment;
• FIG. 5 is a flowchart showing an example of flow of a self-position detection process according to an embodiment;
• FIG. 6 is an illustrative diagram for illustrating a feature point set on an object;
• FIG. 7 is an illustrative diagram for illustrating addition of feature points;
• FIG. 8 is an illustrative diagram for illustrating an example of a prediction model;
• FIG. 9 is an illustrative diagram for illustrating an example of a configuration of feature data;
• FIG. 10 is a flowchart showing an example of flow of an object recognition process according to an embodiment;
• FIG. 11 an illustrative diagram for illustrating an example of procedure data that can be acquired in an embodiment;
• FIG. 12 is a flowchart showing an example of flow of a procedure control process according to an embodiment;
• FIG. 13 is a flowchart showing an example of flow of a procedure displaying process according to an embodiment;
• FIG. 14 is an illustrative diagram showing a first example of an output image output by an image processing device according to an embodiment;
• FIG. 15 is an illustrative diagram showing a second example of an output image output by an image processing device according to an embodiment; and
• FIG. 16 is a block diagram showing an example of a hardware configuration of a general-purpose computer.
DETAILED DESCRIPTION OF THE EMBODIMENT(S)
• Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the appended drawings. Note that, in this specification and the appended drawings, structural elements that have substantially the same function and structure are denoted with the same reference numerals, and repeated explanation of these structural elements is omitted.
• Also, the “detailed description of the embodiment(s)” will be described in the following order.
• 1. Overview of Image Processing Device
• 2. Configuration of Image Processing Device according to Embodiment
• • 2-1. Imaging unit
  • 2-2. Environment Map Generating Unit
  • 2-3. Output Image Generating Unit
  • 2-4. Example of Output Image
• 3. Example of Hardware Configuration
• 4. Conclusion
1. OVERVIEW OF IMAGE PROCESSING DEVICE
• First, an overview of an image processing device according to an embodiment of the present invention will be described with reference toFIGS. 1A to 3B.FIGS. 1A and1B show environments1aand1b, for example, in which theimage processing device100 according to an embodiment of the present invention can be used, respectively.
• Referring toFIG. 1A, asink11, a cuttingboard12, abowl13, astove14, amicrowave oven15, and other physical objects are present in theenvironment1a. That is, theenvironment1ais an environment corresponding to a kitchen as a real space where a user prepares food. Meanwhile, referring toFIG. 1B, a table16, chairs17,dishes18,glasses19 and20, and other physical objects are present in anenvironment1b. That is, theenvironment1bis an environment corresponding to a dining room (or guest seating of a restaurant) as a real space where a user dines.
• FIGS. 2A and 2B are illustrative diagrams for illustrating animage processing device100 that may be used inside theenvironment1aor1b, for example, as described above.
• Referring toFIG. 2A, animage processing device100aincluding abody101a, animaging device102aand adisplay device104a, which are mounted to a user, is shown. Thebody101aincludes a central processing unit (CPU) for executing image processing in theimage processing device100, which will be described in detail later, a storage medium, and the like. Theimaging device102ais mounted to a head portion of the user in the same direction as a user's gaze, to image the inside of theenvironment1aor1bto generate a set of input images. Theimage processing device100aexecutes image processing using the set of input images generated by theimaging device102aas an input to generate a set of output images. The output images generated by theimage processing device100aare displayed by thedisplay device104a. Thedisplay device104ais a head mount display mounted to the head portion of the user. Thedisplay device104amay be, for example, a see-through display.
• Referring toFIG. 2B, animage processing device100bsuch as a mobile phone terminal held by the user is shown. Theimage processing device100bis, for example, a mobile phone terminal with an imaging device, and includes an imaging device (e.g., provided at the rear of adisplay device104b) and thedisplay device104b. Theimage processing device100bincludes a CPU for executing image processing in theimage processing device100, which will be described in detail later, a storage medium, and the like.
• In this disclosure, when theimage processing devices100aand100bneed not be discriminated from each other, letters of the reference numerals are omitted to collectively refer to theimage processing devices100aand100bas theimage processing device100. Also, the same applies to theimaging devices102aand102b(imaging device102), thedisplay devices104aand104b(display device104) and other elements.
• FIGS. 3A and 3B are illustrative diagrams showinginput images106aand106b, for example, acquired by theimaging device102 imaging theenvironments1aand1b, respectively.
• Referring toFIG. 3A, a cuttingboard12, astove14, akitchen knife21, acruet22 and other physical objects are shown in theinput image106a. Theinput image106ais the same image as an image shown in the vision of the user preparing food in theenvironment1a. Theimage processing device100 acquires a set of input images including such aninput image106a, for example, using theimaging device102, and superimposes a direction for a cooking procedure on each input image.
• Referring toFIG. 3B, a table16, adish18,glasses19 and20, aknife23,forks24, anapkin25 and other physical objects are shown in aninput image106b. Theinput image106bis the same image as an image shown in the vision of a user dining in theenvironment1b. Theimage processing device100 acquires a set of input images including such aninput image106b, for example, using theimaging device102 and superimposes a direction for table manners during the meal on each input image.
• An example of a configuration of such animage processing device100 will be described in greater detail in the next section.
2. CONFIGURATION OF IMAGE PROCESSING DEVICE ACCORDING TO EMBODIMENT
• FIG. 4 is a block diagram showing an example of a configuration of theimage processing device100 according to an embodiment of the present invention. Referring toFIG. 4, theimage processing device100 includes animaging unit102, an environmentmap generating unit110, and an output image generating unit180.
• [2-1. Imaging Unit]
• Theimaging unit102 may be realized as an imaging device having an imaging element such as a charge coupled device (CCD) or a complementary metal oxide semiconductor (CMOS), for example. Theimaging unit102 may be provided outside theimage processing device100. Theimaging unit102 outputs an image acquired by imaging the real space such as theenvironment1aor theenvironment1bto the environmentmap generating unit110 and the output image generating unit180 as the input image.
• [2-2. Environment Map Generating Unit]
• The environmentmap generating unit110 generates an environment map representing, for example, positions of one or more physical objects present in the real space based on the input image input from theimaging unit102 and feature data of an object, which will be described later, stored in adata storage unit130. As shown inFIG. 4, in this embodiment, the environmentmap generating unit110 includes a self-position detecting unit120, thedata storage unit130, animage recognizing unit140, an environmentmap building unit150 and an environmentmap storage unit152.
• (1) Self-Position Detection Unit
• The self-position detecting unit120 dynamically detects a position of the imaging device, which takes the input image, based on the input image input from theimaging unit102 and the feature data stored in thedata storage unit130. For example, even in a case in which the imaging device has a monocular camera, the self-position detecting unit120 may dynamically determine a position and posture of the camera and a position of a feature point (FP) on an imaging plane of the camera for each frame by applying the SLAM technology disclosed in Andrew J. Davison's “Real-Time Simultaneous Localization and Mapping with a Single Camera,” Proceedings of the 9th IEEE International Conference onComputer Vision Volume 2, 2003, pp. 1403-1410.
• First, entire flow of a self-position detection process in the self-position detecting unit120 to which the SLAM technology is applied will be described with reference toFIG. 5. Next, the self-position detection process will be described in detail with reference toFIGS. 6 to 8.
• FIG. 5 is a flowchart showing an example of the flow of the self-position detection process in the self-position detecting unit120 to which the SLAM technology is applied. InFIG. 5, when the self-position detection process starts, the self-position detecting unit120 first initializes a state variable (step S102). In this embodiment, the state variable is a vector including the position and the posture (rotation angle) of the camera, a moving speed and an angular speed of the camera and the position of one or more FPs as elements. The self-position detecting unit120 then sequentially obtains the input image from the imaging unit102 (step S112). The process from step S112 to step S118 may be repeated for each input image (i.e., each frame).
• In step S114, the self-position detecting unit120 tracks FPs present in the input image. For example, the self-position detecting unit120 detects a patch (small image of 3×3=9 pixels around a FP, for example) of each FP stored in advance in thedata storage unit130 from the input image. The position of the patch detected herein, that is, the position of the FP, is used to update the state variable later.
• In step S116, the self-position detecting unit120 generates, for example, a predicted value of the state variable of a next frame based on a given prediction model. Also, in step S118, the self-position detecting unit120 updates the state variable using the predicted value of the state variable generated in step S116 and an observed value according to the position of the FP detected in step S114. The self-position detecting unit120 executes the process in steps S116 and S118 based on a principle of an extended Kalman filter.
• As a result of such a process, a value of the state variable updated for each frame is output. Hereinafter, contents of respective processes of tracking the FP (step S114), prediction of the state variable (step S116) and updating the state variable (step S118) will be described more specifically.
• (1-1) Tracking of FP
• In this embodiment, thedata storage unit130 stores the feature data indicating features of objects corresponding to physical objects which may be present in the real space, in advance. The feature data includes small images, that is, the patches regarding one or more FPs, each representing the feature of appearance of each object, for example. The patch may be the small image composed of 3×3=9 pixels around the FP, for example.
• FIG. 6 shows two examples of the objects and an example of FPs and patches set on each object. A left object inFIG. 6 is the object representing a drawer (seeFIG. 6a). A plurality of FPs including a feature point FP1 are set on the object. Further, a patch Pth1 is defined to be associated with the feature point FP1. On the other hand, a right object inFIG. 6 is the object representing a calendar (seeFIG. 6b). A plurality of FPs including a feature point FP2 are set on the object. Further, a patch Pth2 is defined to be associated with the feature point FP2.
• When the input image is acquired from theimaging unit102, the self-position detecting unit120 matches partial images included in the input image against the patch for each FP illustrated inFIG. 6 stored in advance in thedata storage unit130. The self-position detecting unit120 then specifies a position of each FP included in the input image (a position of a center pixel of the detected patch, for example) as the result of matching.
• Further, in tracking of the FPs (step S114 inFIG. 5), data regarding all the FPs to be tracked may not be stored in thedata storage unit130 in advance. For example, six FPs are detected in the input image at time T=t−1 in an example ofFIG. 7 (seeFIG. 7a). Next, when the position or the posture of the camera changes at time T=t, only two of the six FPs present in the input image at the time T=t−1 are present in the input image. In this case, the self-position detecting unit120 may newly set FPs in positions where a characteristic pixel pattern of the input image is present and use the new FPs in the self-position detection process for a subsequent frame. For example, in the example shown inFIG. 7, four new FPs are set on the object at the time T=t (seeFIG. 7b). This is a characteristic of the SLAM technology, and accordingly, a cost of setting all of the FPs in advance can be reduced and accuracy of the process can be improved using a number of added FPs.
• (1-2) Prediction of State Variable
• In this embodiment, the self-position detecting unit120 uses a state variable X shown in the following equation as the state variable to be applied with the extended Kalman filter.
• $\begin{array}{cc}\left[\mathrm{Equation}1\right]& \\ X=\left(\begin{array}{c}x\\ \omega \\ \begin{array}{c}\stackrel{.}{x}\\ \stackrel{.}{\omega }\end{array}\\ {\mathrm{<figure-callout\; label="p"\; filenames="US20110187746A1-20110804-D00008.png,US20110187746A1-20110804-D00009.png"\; state="\{\{state\}\}">p</figure-callout>}}_{1}\\ ⋮\\ p_N\end{array}\right)& \left(1\right)\end{array}$
• The first element of the state variable X in Equation (1) represents a three-dimensional position of the camera in a global coordinate system (x, y, z) being a coordinate system set in the real space, as in the following equation.
• $\begin{array}{cc}\left[\mathrm{Equation}2\right]& \\ x=\left(\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right)& \left(2\right)\end{array}$
• Also, the second element of the state variable is a four-dimensional vector ω having a quaternion as an element corresponding to a rotation matrix representing the posture of the camera. The posture of the camera may be represented using an Euler angle in place of the quaternion. Also, the third and the fourth elements of the state variable represent the moving speed and the angular speed of the camera, respectively.
• Further, the fifth and subsequent elements of the state variable represent a three-dimensional position p_iof a FP FP_i(i=1 . . . N) in the global coordinate system as shown in a following equation. Further, as described above, the number N of the FPs may change during the process.
• $\begin{array}{cc}\left[\mathrm{Equation}3\right]& \\ p_i=\left(\begin{array}{c}{x}_{i}\\ y_i\\ z_{i}\end{array}\right)& \left(3\right)\end{array}$
• The self-position detecting unit120 generates the predicted value of the state variable for a latest frame based on the value of the state variable X initialized in step S102 or the value of the state variable X updated in a previous frame. The predicted value of the state variable is generated according to a state equation of the extended Kalman filter according to multi-dimensional normal distribution shown in the following equation.
• 
  [Equation 4]
• 
  predicted state variable{circumflex over (X)}=F(X,a)+w  (4)
• Here, F denotes the prediction model regarding state transition of a system. a denotes a prediction condition. Also, w denotes Gaussian noise and may include a model approximation error, an observation error and the like, for example. In general, an average of the Gaussian noise w is 0.
• FIG. 8 is an illustrative diagram for illustrating an example of the prediction model according to this embodiment. Referring toFIG. 8, two prediction conditions in the prediction model according to this embodiment are shown. First, as a first condition, it is assumed that the three-dimensional position of the FP in the global coordinate system does not change. That is, provided that the three-dimensional position of the FP FP1 at the time T is p_T, the following relationship is satisfied.
• 
  [Equation 5]
• 
  p_t=p_t-1  (5)
• Next, as a second condition, it is assumed that motion of the camera is uniform motion. That is, a following relationship is satisfied for the speed and the angular speed of the camera from the time T=t−1 to the time T=t.
• 
  [Equation 6]
• 
  {dot over (x)}_t={dot over (x)}_t-1  (6)
• 
  {dot over (ω)}_t={dot over (ω)}_t-1  (7)
• The self-position detecting unit120 generates the predicted value of the state variable for the latest frame based on such a prediction model and the state equation expressed in Equation (4).
• (1-3) Updating of State Variable
• The self-position detecting unit120 then evaluates an error between observation information predicted from the predicted value of the state variable and actual observation information obtained as a result of FP tracking, using an observation equation, for example. Note that v in Equation (8) is the error.
• 
  [Equation 7]
• 
  observation informations=H({circumflex over (X)})+v  (8)
• 
  predicted observation informationŝ=H({circumflex over (X)})  (9)
• Here, H represents an observation model. For example, a position of the FP FP; on the imaging plane (u-v plane) is defined as expressed in the following equation.
• $\begin{array}{cc}\left[\mathrm{Equation}8\right]& \\ \mathrm{position}\mathrm{of}{\mathrm{FP}}_i\mathrm{on}\mathrm{imaging}\mathrm{plane}{\tilde{p}}_i=\left(\begin{array}{c}{u}_i\\ v_i\\ 1\end{array}\right)& \left(10\right)\end{array}$
• Here, all of the position of the camera x, the posture of the camera ω and the three-dimensional position p; of the FP FP_iare given as the elements of the state variable X. Then, the position of the FP FP_ion the imaging plane is derived using the following equation according to a pinhole model.
• 
  [Equation 9]
• 
  λ{tilde over (p)}_i=AR_ω(p_i−x)  (11)
• Herein, λ represents a parameter for normalization, A represents a camera internal parameter, R_ω represents the rotation matrix corresponding to the quaternion ω representing the posture of the camera included in the state variable X. The camera internal parameter A is given in advance as expressed in the following equation according to characteristics of the imaging device, which takes the input image.
• $\begin{array}{cc}\left[\mathrm{Equation}10\right]& \\ A=\left(\begin{array}{ccc}-f·k_u& f·k_u·\cot \theta & u_o\\ 0& -\frac{f·k_v}{\sin \theta }& v_o\\ 0& 0& 1\end{array}\right)& \left(12\right)\end{array}$
• Herein, f represents focal distance, θ represents orthogonality of an image axis (ideal value is 90 degrees), k_urepresents a scale along a vertical axis of the imaging plane (rate of change of scale from the global coordinate system to the coordinate system of the imaging plane), k_vrepresents a scale along a horizontal axis of the imaging plane, and (u_o, v_o) represents a center position of the imaging plane.
• Therefore, a feasible latest state variable X may be obtained by searching the state variable X, which makes the error between the predicted observation information derived using Equation (11), that is, the position of each FP on the imaging plane and the result of FP tracking in step S114 inFIG. 5, minimum.
• 
  [Equation 11]
• 
  lastest state variableX←{circumflex over (X)}+Innov(s−ŝ)  (13)
• The self-position detecting unit120 outputs the position x and the posture ω of the camera (imaging device) dynamically updated by applying the SLAM technology in this manner to the environmentmap building unit150 and the output image generating unit180.
• (2) Data Storage Unit
• Thedata storage unit130 stores in advance the feature data indicating the feature of the object corresponding to the physical object, which may be present in the real space, using a storage medium such as a hard disk or a semiconductor memory. Although an example in which thedata storage unit130 is a part of the environmentmap generating unit110 is shown inFIG. 4, the present invention is not limited to such an example, and thedata storage unit130 may be provided outside the environmentmap generating unit110.FIG. 9 is an illustrative diagram for illustrating an example of a configuration of the feature data.
• Referring toFIG. 9, feature data FD1 is shown as an example for the object Obj1. The feature data FD1 includes an object name FD11, image data FD12 taken from six directions, patch data FD13, three-dimensional shape data FD14 and ontology data FD15.
• The object name FD11 is the name by which a corresponding object may be specified such as a “coffee cup A.”
• The image data FD12 includes six image data obtained by taking images of the corresponding object from six directions: front, back, left, right, above and below, for example. The patch data FD13 is a set of small images around each FP for each of one or more FPs set on each object. The image data FD12 and the patch data FD13 may be used for an object recognition process in theimage recognizing unit140, which will be described later. Also, the patch data FD13 may be used for the above-described self-position detection process in the self-position detecting unit120.
• The three-dimensional shape data FD14 includes polygon information for recognizing a shape of the corresponding object and three-dimensional positional information of FPs. The three-dimensional shape data FD14 may be used for an environment map building process in the environmentmap building unit150, which will be described later.
• The ontology data FD15 is the data that may be used to support the environment map building process in the environmentmap building unit150, for example. In the example ofFIG. 9, the ontology data FD15 indicates that the object Obj1, which is the coffee cup, is more likely to come in contact with an object corresponding to a desk or a dishwasher and is less likely to come in contact with an object corresponding to a bookshelf.
• (3) Image Recognizing Unit
• Theimage recognizing unit140 specifies objects to which physical objects present in the input image correspond, using the above-described feature data stored in thedata storage unit130.
• FIG. 10 is a flowchart showing an example of flow of the object recognition process in theimage recognizing unit140. Referring toFIG. 10, first, theimage recognizing unit140 acquires the input image from the imaging unit102 (step S212). Next, theimage recognizing unit140 matches partial images included in the input image against patches of one or more FPs of each object included in the feature data to extract FPs included in the input image (step S214). The FPs used in the object recognition process in theimage recognizing unit140 and the FPs used in the self-position detection process in the self-position detecting unit120 are not necessarily the same. However, when common FPs are used in the both processes, theimage recognizing unit140 may reuse the result of FP tracking by the self-position detecting unit120.
• Next, theimage recognizing unit140 specifies the object present in the input image based on the result of extracting the FP (step S216). For example, when the FPs belonging to one object are extracted with high density in a certain area, theimage recognizing unit140 may recognize that the object is present in the area. Theimage recognizing unit140 outputs the object name (or an identifier) of the specified object and the position of the FP belonging to the object on the imaging plane to the environment map building unit150 (step S218).
• (4) Environment Map Building Unit
• The environmentmap building unit150 generates the environment map using the position and the posture of the camera input from the self-position detecting unit120, the positions of the FPs on the imaging plane input from theimage recognizing unit140 and the feature data stored in thedata storage unit130. In this disclosure, the environment map is a set of data indicating positions (and postures) of one or more objects present in the real space. The environment map may include object names corresponding to objects, the three-dimensional positions of FPs belonging to objects and the polygon information configuring shapes of objects, for example. The environment map may be built by obtaining the three-dimensional position of each FP according to the above-described pinhole model from the position of the FP on the imaging plane input from theimage recognizing unit140, for example.
• By changing the relation equation of the pinhole model expressed in Equation (11), the three-dimensional position pi of the FP FP_iin the global coordinate system may be obtained by the following equation.
• $\begin{array}{cc}\left[\mathrm{Equation}12\right]& \\ p_i=x+\lambda ·R_{\omega }^T·A^{-1}·{\tilde{p}}_i=x+d·R_{\omega }^T\frac{A^{-1}·{\tilde{p}}_i}{A^{-1}·{\tilde{p}}_i}& \left(14\right)\end{array}$
• Herein, d denotes a distance between the camera and each FP in the global coordinate system. The environmentmap building unit150 may calculate such a distance d based on the positions of at least four FPs on the imaging plane and the distance between the FPs for each object. The distance between the FPs is stored in advance in thedata storage unit130 as the three-dimensional shape data FD14 included in the feature data described with reference toFIG. 9. It should be noted that a calculation process of the distance d in Equation (14) is disclosed in detail in Japanese Patent Application Laid-Open Publication No. 2008-304268
• After the distance d is calculated, remaining variables of a right side of Equation (14) are the position and the posture of the camera input from the self-position detecting unit120 and the position of the FP on the imaging plane input from theimage recognizing unit140, all of which are known. The environmentmap building unit150 then calculates the three-dimensional position in the global coordinate system for each FP input from theimage recognizing unit140 according to Equation (14). The environmentmap building unit150 then builds a latest environment map according to the three-dimensional position of each calculated FP and allows the environmentmap storage unit152 to store the built environment map.
• It should be noted that, at that time, the environmentmap building unit150 may improve accuracy of the data of the environment map using the ontology data FD15 included in the feature data described with reference toFIG. 9.
• The environmentmap storage unit152 stores the environment map built by the environmentmap building unit150 using the storage medium such as the hard disk or the semiconductor memory.
• [2-3. Output Image Generating Unit]
• The output image generating unit180 generates, for a set of procedures of the operation to be performed in the real space, an output image for presenting the direction for each procedure to the user and displays the generated output image on thedisplay device104. In this case, the output image generating unit180 associates a given position in the environment map generated by the environmentmap generating unit110 with the direction for each procedure, and superimposes the direction for each procedure at a position in the input image corresponding to the given position. As shown inFIG. 4, in the present embodiment, the output image generating unit180 includes aprocedure storage unit184, aprocedure control unit186, aninformation arrangement unit190, and asuperimposing unit192.
• (1) Procedure Storage Unit
• Theprocedure storage unit184 stores procedure data defined by causing the direction for each procedure to correspond to the position information designating a position at which the direction is to be displayed, for a set of procedures of operation to be performed in the real space, using a storage medium such as hard disk or a semiconductor memory.
• FIG. 11 is an illustrative diagram for illustratingprocedure data182, for example, stored in theprocedure storage unit184 according to the present embodiment. Referring toFIG. 11, theprocedure data182 has five data items such as “name of directions,” “procedure ID,” “procedure direction,” “related physical object” and “progress condition.”
• The “name of directions” is a name assigned to each operation to be performed in the real space. Procedure data about a set of procedures for one operation can be acquired by designating one name of directions. In the example ofFIG. 11, procedure data of two operations: “table manners” and “frying eggs” is included in theprocedure data182.
• The “procedure ID” is an identifier for identifying each procedure included in the set of procedures. In the example ofFIG. 11, the procedure data for the “table manners” includes three procedures of procedure IDs: “P101,” “P102” and “P103,” Also, the procedure data for “frying eggs” includes four procedures of procedure IDs: “P201,” “P202,” “P203” and “P204.”
• The “procedure direction” is a character string indicating the direction for each procedure to be present to the user. In the example ofFIG. 11, the procedure direction of the procedure of procedure ID=″P101″ of the “table manners” (hereinafter, referred to as procedure P101) is a character string: “sit from a left side of a chair.” The procedure direction of procedure P102 is a character string: “order food and drink.” The procedure direction of procedure P103 is a character string: “set a napkin on your lap.” Also, the procedure direction of procedure P201 of “frying eggs” is a character string: “divide and stir eggs.” The procedure direction of the procedure P202 is a character string: “move the frying pan to the stove.” The procedure direction of the procedure P203 is a character string: “put the eggs onto the frying pan.” The procedure direction of procedure P204 is a character string: “sprinkle salt and pepper.”
• The “related physical object” is position information designating a position at which each procedure direction is to be displayed, using an object name for specifying a physical object associated with each procedure direction. For example, in the present embodiment, each procedure direction is displayed in the vicinity of the physical object specified by the “related physical object.” In the example ofFIG. 11, the related physical object of procedure P101 of the “table manners” is a “chair.” That is, the procedure direction of procedure P101 can be displayed in the vicinity of the chair included in the environment map. The related physical object of procedure P102 is a “menu.” That is, the procedure direction of the procedure P102 can be displayed in the vicinity of the menu included in the environment map. The related physical object of the procedure P103 is a “napkin.” That is, the procedure direction of the procedure P103 can be displayed in the vicinity of the napkin included in the environment map. Further, for example, the “related physical object” for a procedure not associated with any physical object is blank and the procedure direction of this procedure may be displayed in a specific position such as a center of the screen.
• The “progress condition” indicates a condition for progressing a displaying process from each procedure to a next procedure when theimage processing device100 displays a set of procedures. That is, the progress condition may be said to be information designating timing when each procedure direction is to be displayed. In the example ofFIG. 11, the progress condition is defined according any one or combination of the following three patterns.
• First pattern: a state of the physical object in the environment map
• Second pattern: passage of a given time
• Third pattern: an external event
• The first pattern is a pattern in which the progress condition is satisfied (i.e., the displaying process proceeds to a next procedure) when the position or posture of the physical object represented by the environment map is in a given state. For example, the progress condition of the procedure P102 of theprocedure data182 is that “a menu is put down.” Such a progress condition may be satisfied, for example, when the menu is not present on the table in the environment map. Also, for example, the progress condition of the procedure P103 of theprocedure data182 is that “the napkin is on the chair.” Such a progress condition may be satisfied, for example, when in the environment map when the napkin moves from on the table to over the chair (e.g., on the lap of the user seated on the chair).
• The second pattern is a pattern in which the progress condition is satisfied when a given time has elapsed from displaying of a preceding procedure direction or generation of another event. For example, the progress condition of the procedure P201 of theprocedure data182 is a “passage of one minute.” Such a progress condition may be satisfied when one minute has elapsed after the procedure direction of the procedure P201 is displayed. Also, for example, the progress condition of the procedure P203 of theprocedure data182 is a “passage of 30 seconds from stove ignition.” Such a progress condition may be satisfied, for example, when 30 seconds have elapsed from a stove ignition event notified of from an external device (stove).
• The third pattern is a pattern in which the progress condition is satisfied when there has been a given event notification from another device. For example, the progress condition of the procedure P202 of theprocedure data182 is that “the frying pan is on the stove.” Such a progress condition may be satisfied, for example, when the event is notified of from the stove having detected the frying pan being on the stove, for example, using a pressure sensor. Further, the satisfaction of the condition: “there is a frying pan on a stove” may be determined from a state of the physical object in the environment map (in this case, which is the first pattern). An event notification from another device is not limited to such an example, but may be, for example, a notification of the image recognition result (e.g., the result of recognizing a state of a user's hand) from the imaging device, a notification about a user manipulation from an electronic device, a notification about signal reception from a communication device, or the like.
• Theprocedure storage unit184 storessuch procedure data182 as an example, and outputs at least part of theprocedure data182 according to a request from theprocedure control unit186, which will be described later. Further, in the example ofFIG. 4, theprocedure storage unit184 is provided inside theimage processing device100. However, the prevent invention is not limited to such an example, but the procedure data may be stored in a storage medium external to theimage processing device100. In this case, theimage processing device100 may selectively acquire necessary procedure data from the external storage medium, for example, according to an instruction of the user. Also, a plurality of procedures whose states transition from each other as in a state machine may be defined as procedure data instead of a sequential procedure such as theprocedure data182 illustrated inFIG. 11.
• (2) Procedure Control Unit
• Theprocedure control unit186 acquires procedure data for a desired name of directions from theprocedure storage unit184, for example, according to an instruction from the user, and controls displaying of the procedure direction for each procedure included in the set of procedures according to the acquired procedure data.
• FIG. 12 is a flowchart showing an example of flow of a procedure control process in theprocedure control unit186 according to the present embodiment. Referring toFIG. 12, first, theprocedure control unit186 acquires procedure data for a desired name of directions from the procedure storage unit184 (step S302). For example, when the user designates “table manners” as the name of directions, theprocedure control unit186 acquires procedure data for the “table manners” including the set of procedures P101, P102, P103, . . . illustrated inFIG. 11.
• Next, theprocedure control unit186 reads a first procedure from the acquired procedure data (step S304). For example, when the procedure data including the set of procedures P101, P102, P103 . . . illustrated inFIG. 11 is acquired, theprocedure control unit186 initially reads a record of the procedure P101.
• Next, theprocedure control unit186 displays a procedure direction in the vicinity of a related physical object corresponding to the read procedure (step S306). More specifically, theprocedure control unit186 specifies, for example, in the environment map, a position of the related physical object corresponding to the read procedure, and determines the vicinity of the position of the related physical object as a three-dimensional position at which the procedure direction is to be displayed. Theprocedure control unit186 outputs the procedure direction and the three-dimensional position at which the procedure direction is to be displayed, to theinformation arrangement unit190. A subsequent procedure displaying process will be described in greater detail later.
• Next, theprocedure control unit186 monitors a state of the physical object, passage of a given time, an external event, or the like in the environment map according to the progress condition corresponding to the read procedure (step S308). When the progress condition is satisfied as a result of monitoring, the process proceeds to step S310.
• Next, theprocedure control unit186 determines whether a next procedure that has not been displayed remains (step S310). Here, when the next procedure that has not been displayed remains, the process returns to step S304 and the procedure displaying process is repeated for a next procedure. On the other hand, when the next procedure that has not been displayed does not remain, the procedure control process in theprocedure control unit186 ends.
• (4) Information Arrangement Unit
• Theinformation arrangement unit190 calculates a position in the input image at which each procedure direction input from theprocedure control unit186 is to be displayed, according toEquation 11 of the pinhole model using the position and posture of the imaging device acquired from the environmentmap generating unit110. In this case, the three-dimensional position p_iof the FP FP_iat the right side of Equation (11) is substituted with the three-dimensional position input from theprocedure control unit186. After theinformation arrangement unit190 calculates the position in the input image at which each procedure direction is to be displayed, theinformation arrangement unit190 outputs each procedure direction and the position in the input image at which the procedure direction is to be displayed, to thesuperimposing unit192.
• (5) Superimposing Unit
• The superimposingunit192 generates an output image by superimposing each procedure direction input from theinformation arrangement unit190 at the position in the input image calculated by theinformation arrangement unit190.
• FIG. 13 is a flowchart showing an example of flow of a procedure displaying process in theprocedure control unit186, theinformation arrangement unit190 and the superimposingunit192. Further, the procedure displaying process shown inFIG. 13 is executed in step S306 in the procedure control process shown inFIG. 12 for each individual procedure included in a set of procedures of operation to be performed in the real space.
• Referring toFIG. 13, first, theprocedure control unit186 determines the three-dimensional position at which the procedure direction is to be displayed, based on a position of a related physical object in the environment map (step S322). For example, when the related physical object for any procedure is on a chair, a position of either a surface or the vicinity of the chair in the environment map is determined as the three-dimensional position at which the procedure direction for the procedure is to be displayed.
• Next, theinformation arrangement unit190 calculates the position in the input image corresponding to the three-dimensional position determined in step S322 according to the pinhole model using the position and posture of the imaging device acquired from the environment map generating unit110 (step S324). Further, when the position on the imaging plane corresponding to the three-dimensional position determined in step S322 is out of a range of the input image, a subsequent process may be skipped.
• Next, the superimposingunit192 generates an output image by superimposing, for example, a text box describing the procedure direction at the position in the input image calculated by the information arrangement unit190 (step S326). The output image generated by the superimposingunit192 is displayed, for example, on the screen of thedisplay device104 of theimage processing device100.
• [2-4. Example of Output Image]
• FIGS. 14 and 15 show examples of an output image that can be displayed on the screen of thedisplay device104 in the present embodiment, respectively.
• Referring toFIG. 14, an output image Im11 on which the procedure direction for the procedure P202 illustrated inFIG. 11 is superimposed is shown. In the output image Im11, a text box T11 describing a procedure direction “procedure2: move the frying pan to the stove” is displayed in the vicinity of the frying pan. The user can intuitively recognize that it is good for the frying pan to be on the stove in a next cooking procedure by viewing such a procedure direction. This text box T11 continues to be displayed in the vicinity of the frying pan until a state “the frying pan is on the stove,” which is the progress condition of the procedure P202, is realized in the environment map. Thus, the user can easily recognize a physical object serving as an object of the procedure.
• Referring toFIG. 15, an output image Im21 on which a procedure direction for the procedure P103 illustrated inFIG. 11 is superimposed is shown. In the output image Im21, a text box T21 describing a procedure direction “procedure3: set a napkin in your lap” is displayed in the vicinity of the napkin. The user can intuitively recognize that it is good for the napkin to be set on the lap in a next procedure of the table manners by viewing such a procedure direction. This text box T21 continues to be displayed in the vicinity of the napkin until a state “the napkin is on the chair,” which is the progress condition of the procedure P103, is realized in the environment map. Thus, the user can easily recognize a physical object serving as an object of the procedure.
• The procedure display method described herein is only an example, and various applications are possible, in addition to the above-described embodiment. For example, the present invention may be applied for directions in spaces such as commercial facilities or train stations or for manipulation directions in a rent-a-car or an automated teller machine (ATM).
3. EXAMPLE OF HARDWARE CONFIGURATION
• It does not matter whether a set of processes according to the above-described embodiment are realized by hardware or software. When a set of processes or a part of the same is executed by the software, a program composing the software is executed using a computer incorporated in dedicated hardware or a general-purpose computer shown inFIG. 16, for example.
• InFIG. 16, theCPU902 controls entire operation of the general-purpose computer. A program or data describing some or all of the processes in the set is stored in a read only memory (ROM)904. The program and data used by theCPU902 in process execution are temporarily stored in a random access memory (RAM)906.
• TheCPU902, theROM904, and theRAM906 are connected to each other via abus910. Further, an input/output interface912 is connected to thebus910.
• The input/output interface912 is an interface for connecting theCPU902, theROM904, and theRAM906 with theinput device920, thedisplay device104, thestorage device924, theimaging device102, and thedrive930.
• Theinput device920 receives instructions or information input from the user, for example, via an input interface such as a button, a switch, a lever, a mouse, or a keyboard. Thestorage device924 includes, for example, a hard disk drive or a semiconductor memory, and stores programs or data. Thedrive930 is provided in the general-purpose computer as necessary and, for example, aremovable media932 is mounted in thedrive930.
• When the set of processes is executed by software, for example, a program stored in theROM904, thestorage device924, or theremovable media932 shown inFIG. 16, when executed, is read into theRAM906 and then executed by theCPU902.
4. CONCLUSION
• The embodiment of the present invention has been described with reference toFIGS. 1A to 16. According to the present embodiment, the direction for each procedure included in the set of procedures of the operation to be performed in the real space is displayed at the position determined based on the environment map, which three-dimensionally represents a position of a physical object present in the real space. Accordingly, the procedure can be intuitively understood by the user. In particular, according to the present embodiment, the environment map is dynamically updated to follow a change in the environment even in the operation in the real space in which an environment surrounding the user may be dynamically changed. As a result, the position at which the direction for each procedure is displayed moves according to the change of the environment, thereby preventing difficulty in understanding of the directions from being caused due to the change of the environment.
• Also, in the present embodiment, the position at which the direction for each procedure is to be displayed can be determined according to the position of the physical object associated with the direction for each procedure. Accordingly, it is possible for the user to easily recognize the physical object as a target object of the operation in each procedure. Also, since the timing when the direction for each procedure is to be displayed is controlled according to the state (position or posture) of the physical object in the environment map, passage of a given time, an external event or the like, the direction can be displayed at proper timing according to the progress of the operation by the user. Further, the position at which the direction for each procedure is to be displayed is calculated based on the position and posture of the camera dynamically detected using SLAM technology. Accordingly, even when a camera whose position or posture is likely to be changed from moment to moment is used, the direction can be displayed in a proper position in the image.
• The preferred embodiments of the present invention have been described above with reference to the accompanying drawings, whilst the present invention is not limited to the above examples, of course. A person skilled in the art may find various alternations and modifications within the scope of the appended claims, and it should be understood that they will naturally come under the technical scope of the present invention.
• The present application contains subject matter related to that disclosed in Japanese Priority Patent Application JP 2010-021368 filed in the Japan Patent Office on Feb. 2, 2010, the entire content of which is hereby incorporated by reference.

Claims (8)

1. An image processing device comprising:

a data storage unit having feature data stored therein, the feature data indicating a feature of appearance of one or more physical objects;

an environment map building unit for building an environment map based on an input image obtained by imaging a real space using an imaging device and the feature data stored in the data storage unit, the environment map representing a position of a physical object present in the real space;

a control unit for acquiring procedure data for a set of procedures of operation to be performed in the real space, the procedure data defining a correspondence between a direction for each procedure and position information designating a position at which the direction is to be displayed; and

a superimposing unit for generating an output image by superimposing the direction for each procedure included in the set of procedures at a position in the input image determined based on the environment map and the position information, using the procedure data acquired by the control unit.

2. The image processing device according toclaim 1,

wherein the position information designates a position in the environment map at which the direction is to be displayed, by specifying a physical object related to the direction for each procedure.

3. The image processing device according toclaim 1,

wherein the procedure data defines a further correspondence between the direction for each procedure and a condition for progressing displaying of each direction.

4. The image processing device according toclaim 3,

wherein the condition for progressing displaying of each direction includes a condition according to a position or a posture of a physical object represented by the environment map.

5. The image processing device according toclaim 3,

wherein the control unit controls displaying of the direction for each procedure included in the set of procedures according to the procedure data.

6. The image processing device according toclaim 1,

wherein the image processing device further comprises a detection unit for dynamically detecting a position in the real space of the imaging device based on the input image and the feature data, and

the position in the input image at which the direction for each procedure is to be superimposed is determined according to the position in the real space of the imaging device detected by the detection unit.

7. An image processing method in an image processing device including a storage medium having feature data stored therein, the feature data indicating a feature of appearance of one or more physical objects, the method comprising the steps of:

acquiring an input image obtained by imaging a real space using an imaging device;

building an environment map based on the input image and the feature data, the environment map representing a position of a physical object present in the real space;

acquiring procedure data for a set of procedures of operation to be performed in the real space, the procedure data defining a correspondence between a direction for each procedure and position information designating a position at which the direction is to be displayed; and

generating an output image by superimposing the direction for each procedure included in the set of procedures at a position in the input image determined based on the environment map and the position information, using the acquired procedure data.

8. A program for causing a computer, which controls an image processing device including a storage medium having feature data stored therein, the feature data indicating a feature of appearance of one or more physical objects, to function as:

an environment map building unit for building an environment map based on an input image obtained by imaging a real space using an imaging device and the feature data, the environment map representing a position of a physical object present in the real space;

a control unit for acquiring procedure data for a set of procedures of operation to be performed in the real space, the procedure data defining a correspondence between a direction for each procedure and position information designating a position at which the direction is to be displayed; and

a superimposing unit for generating an output image by superimposing the direction for each procedure included in the set of procedures at a position in the input image determined based on the environment map and the position information, using the procedure data acquired by the control unit.

Images